Structural determinants of phosphorylation‐dependent nuclear transport of HCMV DNA polymerase processivity factor UL44

Human cytomegalovirus DNA polymerase processivity factor UL44 is transported into the nucleus by importin (IMP) α/β through a classical nuclear localization signal (NLS), and this region is susceptible to cdc2‐mediated phosphorylation at position T427. Whilst phosphorylation within and close to the UL44 NLS regulates nuclear transport, the details remain elusive, due to the paucity of structural information regarding the role of negatively charged cargo phosphate groups. We addressed this issue by studying the effect of UL44 T427 phosphorylation on interaction with several IMPα isoforms by biochemical and structural approaches. Phosphorylation decreased UL44/IMPα affinity 10‐fold, and a comparative structural analysis of UL44 NLS phosphorylated and non‐phosphorylated peptides complexed with mouse IMPα2 revealed the structural rearrangements responsible for phosphorylation‐dependent inhibition of UL44 nuclear import.

Active nuclear import is a signal and energy-dependent process mediated by karyophilic proteins belonging to the importin (IMP) superfamily, whereby IMPb1 or one of its several homologs recognize nuclear localization signals (NLSs) on the cargo protein, translocate the latter through the nuclear pore complex (NPC), and release it in the nucleoplasm upon binding the small GTPase Ran [1].The best-characterized NLSs are highly basic sequences recognized by IMPb1 through the adapter molecule IMPa [2], and are known as classical (c)NLSs.Human IMPas exist as seven isoforms (hIMPa1/3/4/5/6/7/8), with the mouse homolog of hIMPa1, "mIMPa2", frequently used for structural studies [3][4][5][6][7].cNLSs can either be monopartite or bipartite.Monopartite cNLSs are formed by a single stretch of basic amino acids and bind to the IMPa major binding site, which contains six main binding pockets (P0-P6) able to accommodate NLS aminoacidic side chains.Bipartite cNLSs are formed by two stretches of basic amino acids typically separated by an amino acid linker region and interact with both IMPa major and minor binding sites.The minor site contains four main binding pockets (P1 0 -P4 0 ) and this additional binding interface lowers the K d of the interaction [3,4].A summary of the two binding pockets can be found in Fig. 1.
Nuclear transport of several cargoes is tightly regulated by a number of mechanisms, the best characterized being cargo phosphorylation [8].However, the lack of extensive structural data leaves several unanswered questions [9].Firstly, it is extremely challenging to predict the outcome of cargo phosphorylation on nuclear import.Indeed, cargo phosphorylation can either promote [10][11][12][13][14][15] or impair the process [11,[16][17][18][19][20].Secondly, although phosphorylation-dependent regulation of nuclear import has been proposed to rely on the modulation of cargo-IMP interactions, it remains uncertain whether the latter depends on a change in affinity for IMPs [13,17] or for other factors that compete with IMPs [21 -23].Additionally, phosphorylation of certain cargoes has been proposed to result in conformational changes regulating NLS exposure, and therefore IMP binding and subsequent nuclear import, as seen in hepatitis B core protein where phosphorylation likely acts as an important modulator of the viral life cycle [24].
Nuclear transport of human cytomegalovirus (HCMV) DNA polymerase processivity factor UL44 is mediated by the importin IMPa/b heterodimer, whereby IMPa recognizes a C-terminally located NLS (PNTKKQK-431) and IMPb translocates the complex to the nucleoplasmic side of the NPC [10].Upstream of the NLS, a stretch of serine residues (Fig. 2, indicated as red in the single letter amino acid sequence), has been shown to be the target of a phosphorylation cascade mediated by protein kinases CKII and CKI, which results in enhanced nuclear localization [11,29].
Conversely, residue T427, a target for cdk1/cycBmediated phosphorylation in vitro, lies immediately upstream of the UL44-NLS basic cluster that mediates the interaction with IMPa.The phosphomimic D427 substitution has been shown to reduce both the nuclear import of exogenously expressed GFP-UL44 by live cell confocal laser scanning microscopy and oriLyt-dependent DNA replication [11,29].Although the negative regulator of nuclear import BRAP2 has been shown to decrease nuclear import of the T427D phosphomimic UL44 derivative, the impact of phosphorylation on T427 on UL44 interaction with IMPa remains unknown.We approached this question by biochemically and structurally comparing the binding properties of several IMPa isoforms to an unphosphorylated UL44 NLS peptide (UL44_410-433; 410-KEESDSEDSVTFEFVPNTKKQKCG-433) and its pT427 counterpart (UL44_410-433_pT 410-KEE SDSEDSVTFEFVPNTKKQKCG-433) Fig. 1.Major and minor binding sites of mIMP⍺2.ARM repeats of mouse IMP⍺2 are labeled and color coded (rainbow).The major binding site is found within ARMS 2-4 where the NLS binds at P0-P6 positions, typically with a K-K/R-X-K/R consensus at the P2-P5 sites.The minor binding site is found within ARMS 6-8 where the NLS binds at P1 0 -P4 0 positions, typically with a K-R consensus at the P1 0 and P2 0 sites, respectively.Bipartite NLSs span continuously across both major and minor binding sites with an amino acid linker region between each (not shown) [3,4].(phosphorylated amino acid is underlined).Quantitative fluorescence polarization assays showed that phosphorylation directly affects the ability of UL44 to interact with IMPas, decreasing binding affinity by c. 10-fold.Crystallographic analysis of peptides in complex with mIMPa2 confirmed that UL44 residues 425-431 represent a cNLS, and revealed subtle, but important differences in the IMPa interaction induced by phosphorylation of T427, thus partially explaining the molecular details of phosphorylation-dependent regulation of nuclear import.Despite the main features of the NLS-IMPa interaction being unaffected, with K428, K429, and K431 occupying the key P2, P3, and P5 positions in the mIMPa2 major binding site, a small number of structural changes were observed, including an important conformational change of mIMPa2 R238 sidechain interactions.

Recombinant protein expression and purification
Representatives from each importin (IMP) superfamily (hIMPa3,5,7 and mIMPa2) were recombinantly expressed and purified for assessment of UL44 peptide binding (hIMPa3,5,7 and mIMPa2) and structure determination (mIMPa2).For this, plasmids encoding IMPs with truncated IMPb binding (IBB) domains (first ~70 amino acids removed, summary available in Table S1), and an N-terminal His tag separated by a TEV protease cleavage site, were codon optimized and inserted into pET-30 vectors at BamHI sites (Genscript, Piscataway, NJ, USA).The plasmids were transformed into competent BL21 (DE3) pLysS Escherichia coli cells and expressed for 24-48 h using autoinduction.Cells were harvested and suspended in affinity buffer (50-mM phosphate buffer, 300-mM sodium chloride, 20-mM imidazole, pH 8.0).Before purification, the whole cell was freeze-thawed thrice, then lysed with lysozyme and treated with DNAse.The resulting supernatant containing soluble protein was collected and filtered, injected over a Nickel HisTrap HP 5 mL (Cytvia, Marlborough, MA, USA), washed with affinity buffer, then with elution buffer (50-mM phosphate buffer, 300-mM sodium chloride, 500-mM imidazole, pH 8.0).Peak fractions were collected and treated with TEV protease, incubated at 4 °C overnight (excluding IMPa2 which does not contain a TEV site).The sample was then injected over a HiLoad 26/600 Superdex 200 pg column (Cytvia) using a Tris buffer (50-mM Tris, 125-mM sodium chloride, pH 8.0).Purified importin protein was concentrated using Amicon Ultra-15 Centrifugal Filter Units (Merck, Millipore, Burlington, VT, USA), then aliquoted, and stored at À80 °C until further use in binding and/or crystallization experiments.

Synthesis of UL44 peptides
The UL44_410-433 and UL44_410-433_pT peptides were chemically synthesized employing a solid-phase technique performed on a fully automated peptide synthesizer (Syro II; MultiSynTechGmbh, Witten, Germany).Wang resins preloaded with the first N-a-Fmoc-protected amino acid were employed for stepwise assembly of the entire peptide chain.This assembly was performed according to the Fmoc standard strategy and was based on the use of HATU as the coupling reagent [30].The side-chain protected amino acid building blocks were: yloxycarbonyl-L-lysine, and N-a-Fmoc-S-trityl-L-cysteine.A step of deprotection of the final peptides was conducted, followed by cleavage from the resin with a mixture of 88% (v/v) trifluoroacetic acid (TFA) with 5% phenol (w/v), 5% H 2 O (v/v), and 2% (v/v) of triisopropylsilane via shaking at RT for 2.5 h.A step of vacuum filtration permitted the removal of the resin from the assembled peptide chains.Then, the peptides were precipitated with cold diethyl ether and transformed into pellets by a centrifugation procedure.Two washes with cold diethyl ether were performed on the precipitated peptides.The following purification steps were performed through flash chromatography (SP1; Biotage, Uppsala, Sweden) on a Biotage SNAP Ultra C18 12-g cartridge packed with Biotage HP-Sphere C18 25 lm spherical silica.A final step of molecular mass confirmation was performed by mass spectroscopy on a MALDI-TOF/TOF mass spectrometer (ABI 4800; AB Sciex, Framingham, MA, USA).N-terminally FITC-tagged peptide UL44_410-433 corresponding to the HCMV C-terminal 24 residues (410-KEESDSEDSVTFEFVPNTKKQKCG-433) and its phosphorylated variant UL44_410-433_pT (410-KEESDSE DSVTFEFVPNTKKQKCG-433) were synthesized by the Dementia Research Centre, Macquarie University and used for fluorescence polarization assays as described in Ref. [6].

Fluorescence polarization assays
To quantify the effect of T427 phosphorylation on the binding affinity of UL44 peptides and mIMPa2/ hIMPa3,5,7, FP assays were performed using FITC-tagged NLS peptides and bacterially expressed IMPaDIBBs.Two-fold dilutions of 25 lM IMPa were titrated across 11 wells of a black Fluotrac microplate (Greiner Bio-One, Kremsm€ unster, Austria) and 100-nM FITC peptide was added to each well.The final volume per well was 200 lL using tris-buffered saline pH 8.0.A control containing no IMPa was also prepared, and this was used for gain adjustment.FP measurements were immediately recorded using a CLARIOstar Plus plate reader (BMG Labtech, Ortenberg, Germany) and the assay was repeated in triplicate for each peptide.Binding curves were plotted using nonlinear regression one site-specific binding in GRAPHPAD PRISM (Version 9.5.1):GraphPad Software, Boston, MA, USA.
Crystallization of mIMPa2 and UL44 NLSs mIMPa2 and peptides UL44_410-433 or UL44_410-433_pT were co-crystallized for assessment of their molecular binding interfaces.For this, purified mIMPa2DIBB and unlabeled peptides were mixed in a 3 : 1 molar ratio and screened using hanging drop vapor diffusion techniques in known conditions (tri-sodium citrate 500-850 mM, 0.1 M HEPES pH 6.5 or 7.0 or 7.5, 0.01 M DTT) at 23 °C.Drop size was 3 lL of 1 : 1 protein/peptide mix : reservoir solution suspended over 300 lL of the screening reservoir condition.High-quality crystals used for structure determination of mIMPa2:UL44_410-433 were formed in tri-sodium citrate 0.65 M, HEPES pH 7.5, DTT 0.01 M. For mIMPa2: UL44_410-433_pT, crystals grew in tri-sodium citrate 0.65 M, HEPES pH 7.0, DTT 0.01 M. Large, thick rod-like crystals were harvested in CryoLoops (Hampton Research, Aliso Viejo, CA, USA), cryo-protected using the screen condition supplemented with 20% glycerol, and flash cooled in liquid nitrogen.

Data collection, processing, and structure determination
Crystals were diffracted on the MX2 beamline at the Australian Synchrotron [31].Each dataset contained 3600 images from 360°crystal rotation using 0.1°oscillations.Data space group determination and integration were performed using DIALS [32], then scaled and merged in AIMLESS (CCP4 SUITE, [33]).To guarantee a truly free test set of reflections for bias assessment, a previously solved mIM-Pa2DΙΒΒ model (PDB: 7RG1) was subjected to Cartesian simulated annealing using PHENIX REFINE and the untouched test set was later used together with the output PDB as the initial start model for molecular substitutions.Iterative cycles of model building in COOT [34] and refinement in PHE-NIX REFINE [35]

Phosphorylation of HCMV UL44 T427 reduces binding to IMPa isoforms
We have previously shown that the phosphomimic substitution T427D, immediately upstream of HCMV-UL44 DNA polymerase processivity factor cNLS (426-PNTKKQK-431) strongly inhibits nuclear accumulation, and decreases oriLyt-dependent DNA replication [11].Such reduction in nuclear accumulation was speculated to depend on sequestration by the negative regulator of nuclear import BRAP2, which reduces nuclear import of UL44, but not of the phosphonull substitution derivative T427A [22].Nonetheless, a recent study revealed that phosphorylation in the vicinity of the nuclear localization signal of human dUTPase abolishes nuclear import, and phosphomimic substitution with a negatively charged glutamic acid residue (S11E) at such position decreased affinity of the dUTPase for IMP a [17].We, therefore, hypothesized that phosphorylation of T427 could similarly reduce binding to IMPa.To address this possibility, we employed fluorescence polarization (FP) to quantitatively evaluate the impact of T427 phosphorylation on the binding of FITC-labeled UL44 peptides to several IMPaDIBB isoforms.In such assays, peptide UL44_410-433 bound preferentially to hIMPa5 (K d ~160 nM) as compared to the other IMPa isoforms (K d range 450-750 nM; see Fig. 3).Strikingly, peptide UL44_410-433_pT interacted with all IMPa isoforms with approximately 10-fold lower binding affinity (K d range 1670-10 710 nM).

Phosphorylation of HCMV UL44 T427 changes interactions with mIMPa2DIBB
Since HCMV UL44 peptide UL44_410-433_pT bound IMPa isoforms with a 10-fold reduced affinity as compared to UL44_410-433, it suggested that the negative charge of the phosphate moiety might affect interaction of the NLS basic residues with IMPa binding sites.To address this possibility, we solved the crystal structures of UL44_410-433 and UL44_410-433_pT in complex with mouse IMPa2DIBB to 2.0 A (R work = 19%, R free = 22%) and 1.9 A (R work = 18%, R free = 20%), respectively (Fig. 4).
Peptide NLS residues 425-433 were elucidated for the UL44_410-433 structure, and 421-433 for the UL44_410-433_pT structure.Our results showed that phosphorylation did not markedly alter the binding mode at the key P0-P6 pockets of the IMPa major binding site.Indeed, in both cases, peptides bound to the IMPa major binding site and displayed major site positions like the prototypical Simian Vacuolating virus 40 (SV40) large tumor antigen (LTA) NLS peptide.At the P2 site, a key residue for the NLS, the K428 makes three hydrogen bonds (with the carbonyl group of G150, the hydroxyl group of T155, and the carbonyl oxygen of D192 of IMPa) as well as a salt bridge interaction with the negatively charged IMPa D192.Notably, in the UL44_410-433 peptide, the carbonyl group of P425, accommodated in the P-1 binding pocket (Fig. 4), makes two hydrogen bonds with IMPa R238 (NH1 and NH2).Conversely, in the UL44_410-433_pT structure, IMPa R238 (NH1 and NH2) establishes electrostatic interactions via two hydrogen bonds with phosphorylated threonine (T427) oxygen atoms (OG1, O2P), thereby losing contact with P425.The NLS peptide electron densities and conformational change between the two peptides are highlighted in Fig. 5.
Overall, these results show that there are structural rearrangements of the NLS that occur N terminus to the P2 site lysine that could account for changes in binding affinity observed in our assays.

Discussion
This is one of the first studies to report the crystal structure of a phosphorylated-NLS residue negatively affecting binding to IMPa.Despite cargo phosphorylation being long known to regulate nuclear transport [12], its mechanistic details are still lacking, and at present it is not possible to predict the effect of phosphorylation on nuclear import [9].This is mainly due Fig. 3. Quantitative analysis of the effect of T427 phosphorylation on binding to mIMPa2, hIMPa3, hIMPa5, and hIMPa7 isoforms.(A) FITClabeled peptides (100 nM) were incubated with two-fold serial dilutions of bacterially purified recombinant IMPaDIBBs (range: 25 lM to 2.4 9 10 À2 lM) and subjected to fluorescence polarization (FP) analysis.Data are shown in millipolarized units (mP) as mean AE SEM relative to three independent experiments.Data were analyzed with GRAPHPAD PRISM using the one site-specific binding least squares fit function to calculate the B max and K d relative to the peptide:IMPaDIBBs interaction.(B) A summary table relative to data shown in (A), with mean B max and K d values relative to the interaction of indicated peptides with IMPaDIBBs, along with the respective SEM.  to the lack of structural comparative analysis of phosphorylated and unphosphorylated cargoes in complex with IMPas.However, at least for cNLSs recognized by IMPa, some reports are beginning to shed light on this issue.The majority of information available comes from structural studies with cNLS bearing phosphomimic substitutions in complex with IMPa [17], or molecular dynamics simulation using the structure of IMPa with unphosphorylated cargo peptides [15].To date, the only published study whereby the structure of a phosphorylated residue with IMPa has been reported concerns the complex between EBNA1 cNLS and mIMPa2 [14].The peptide (379-KRPRSPSS-386) could be observed at both the major and minor IMPa binding sites.However, pS385 could only be unambiguously visualized at the minor binding site, thus providing no information regarding the role of S385 phosphorylation on functional interaction at the physiological major IMPa binding site [14].
Overall, it appears that phosphorylation of cargoes upstream of their cNLS can either enhance or decrease nuclear import [11,13,14,37].The mechanism of such regulation is still elusive.In the case of SV40 LTA, the binding has been proposed to rely on conformational changes induced by phosphorylation of S111/112, rather than on the ability of phosphate groups to directly interact with IMPa [11,13,37,38].Alternatively, phosphorylation of residues that lie in close proximity or within the cNLS itself (i.e., positions from P-2 to P1) decreases nuclear import.This has been extensively shown for several viral and cellular proteins, including SV40 LTA, HCMV UL44, human dUTPase, and yeast Swi6 [2,11,16,18].In such cases, reduction in NLS activity has been attributed to reduced affinity for IMPa/b, as well as enhanced affinity to a cytoplasmic retention factor.
Our investigation can establish a connection between the position of the phosphorylated residue in monopartite cNLSs and the mechanism underlying the inhibition of nuclear import.Indeed, in the case of SV40 LTA (124-TPPKKKRKV-132), a negative charge at the P-2 residue T124 does not decrease affinity for IMPa but promotes interaction with the cytoplasmic retention factor BRAP-2 [22].However, for yeast Swi8 (160-SPLKKLKI) and human dUTPase (11-SPSKRARP-18), a negative charge at the P-1 residues decreases binding affinity 5-and 10-fold, respectively [17,18].We show here that phosphorylation at the P1 residue similarly reduces binding of HCMV UL44 NLS to IMPa by 10-fold (see Fig. 3).This suggests that phosphorylation at major site residues P-1/P1 directly decreases IMPa binding, in contrast to what reported for phosphorylation at position P-2.
Further to this, our comparative structural analysis provides valuable mechanistic insights into the process, revealing that phosphorylation at UL44 T427 changes the hydrogen bonding pattern at the mIMPa2 major binding site.Regarding the human dUTPase NLS, a P-1 site phosphomimic displayed structural rearrangement where new contacts with IMPa D270 resulted in the loss of IMPa interaction with the P0 site proline and a conformational change to the P3 site arginine which resulted in a loss of interaction with IMPa Ν228 [17].Phosphorylation of the P1 site in our study did not reveal major conformational changes to the P3 site K429.However, we did see a loss of two hydrogen bonds between the P-1 site P425 and mIMPa2 R238, where the R238 sidechain created two new hydrogen bonds with the phosphorylated threonine.We speculate the loss of interactions at the N-terminal region of the peptide at the major site could weaken the binding in the phosphorylated NLS, similar to what has been proposed for dUTPase.
Hence, the structural differences described here can, to some extent, explain the difference in binding affinity between IMPa and UL44_410-433 and UL44_410-433_pT.Interestingly, it was recently reported that phosphorylation of T88 within the linker of the TDP-48 bipartite NLS reduces IMPa/b binding and nuclear import by affecting the dynamics of NLS conformation.Indeed, MD simulations suggested that CKImediated phosphorylation of the NLS linker reduces the spontaneous transition of the NLS backbone to more kinetically favorable states, decreasing the probability of interaction with IMPa [39].Although the NLS of UL44, Swi6, and dUTPase are monopartite and not bipartite, an effect of N-terminal phosphorylation on the conformation of their NLS cannot be excluded.Finally, it is worth mentioning that SUMOylation of UL44 residue K410, closely located to the phosphorylation-regulated NLS domain, controls UL44 subnuclear distribution and viral replication [28,40,41], and it is, therefore, possible that UL44 SUMOylation and phosphorylation are interconnected, as shown for heat shock transcription factor 4 isoform b, adding another layer of complexity to the regulation of UL44 function and nuclear transport [42].

200FEBS
Letters 598 (2024) 199-209 ª 2023 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

Fig. 2 .
Fig.2.Schematic representation of UL44 functional domains and determinants of phosphorylation regulated nuclear import.The HCMV UL44 sequence is represented by a horizontal white bar.Its catalytic N-terminal domain (residues 1-290) and its highly unstructured Cterminal domain (residues 291-433) are indicated.The vertical orange bars represent residues essential for dimerization (L86, L87, and F121).The vertical blue bar indicates the residue essential for binding to DNA polymerase catalytic subunit pUL54 I135 in the connector loop.Vertical black bars are residues essential for DNA binding (residues 165, 167, and 168 in the gap loop).Vertical gray bars represent glycine-rich stretches.The vertical green bar represents the phosphorylation-dependent nuclear transport region (residues 410-433) and the vertical red bar is the NLS (residues 425-431).The protein sequence encompassing the UL44 phosphorylation region and NLS (residues 410-433) is shown using the single letter amino acid code, with phosphorylation sites enhancing nuclear import depicted in red, and phosphorylation sites decreasing nuclear import depicted in yellow.The NLS is underlined[10,11,[25][26][27][28].

204FEBS
Letters 598 (2024) 199-209 ª 2023 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.
FEBS Letters 598 (2024) 199-209 ª 2023 The Authors.FEBS Letters published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.